[austinjp]

While that's interesting I don't think it's "best evidence".

Unfortunately our mechanisms for determining best evidence are woefully slow, complex, and expensive. Systematic reviews, for instance. Either this process needs to be improved (an excellent candidate for AI) or a nearly-as-good surrogate needs to be supported by policy makers.

[robbiep]

what are your thoughts on how an AI would perform this task?

[austinjp]

My thoughts on this are loosely formed :)

However....

The systematic review and meta-analyses processes involve reading screeds of text and data; organising it; analysing it; weighing it against measures of quality; and conducting (often counter-intuitive) statistical procedures.

This is the sort of thing that takes humans a loooooong time and is prone to error. A domain-specific AI would surely excel at tasks like this.